In this episode, I break down how to run Stable Diffusion efficiently on AMD GPUs that no longer support ROCm, using Vulkan as the acceleration backend. I also cover the full Fedora 43 GPU passthrough fix, where the system continues hiding the GPU even after removing VFIO files from modprobe and rebuilding dracut. This walkthrough is based on the AMD Instinct MI60 32GB HBM2 and the v1-5-pruned-emaonly-fp16.safetensors model running in stable-diffusion.cpp.
Read the full blog article from my earlier tutorial here:
https://www.ojambo.com/review-generative-ai-v1-5-pruned-emaonly-fp16-safetensors-model
Watch the complete long-form video here:
https://youtube.com/live/RhtKQbqiCaQ
Programming Books:
https://www.amazon.com/stores/Edward-Ojambo/author/B0D94QM76N
Programming Courses:
https://ojamboshop.com/product-category/course
One-on-One Programming Tutorials:
https://ojambo.com/contact
AI Installation or Migration (Stable Diffusion, v1-5 fp16 model, etc.):
https://ojamboservices.com/contact
#Vulkan #StableDiffusion #AMD #AMDGPU #Linux #GPUFix #MachineLearning #AIArt #GenerativeAI #Fedora #InstinctMI60
